testing the three vicious circles of the romanian economy: a … · 2017-12-11 · testing the...

Testing the Three Vicious Circles of the RomanianEconomy: A Novel Approach∗

Ioan - Gabriel Stavre†

Doctoral School of Finance and Banking, Bucharest University ofEconomic Studies

October 24, 2017

Abstract

The functioning of fiscal stimulus may adversely affect the normal adjustmentmechanism of the economy, in other words we say we are caught in a vicious circle ifa number of processes are perpetuated in a harmful and reversible way. The aim ofthis paper is to investigate a number of three vicious circles that (presumably) act onthe Romanian economy and have the power to explain (i) the lack of competitivenessof local firms, (ii) the size of the informal economy and the catalyst of the threecircles, namely (iii) the procyclical behavior of fiscal policy. Croitoru & Tarhoaca(1997) were the first to put forward the mechanism driving this disequilibrium.Until now there has been no work attempting to validate the presence of the viciouscircles using a quantitative assessment. We tackle this challenge head on: we use aStructural Vector Autoregressive model using proxy variables, identified using signrestrictions derived from the aforementioned paper and analyze the response of theunrestricted variables. Our results do not point in the direction of the authors, thuswe cannot confirm the presence of these circles in our economy (at least with thecurrent methodology).

Keywords: vicious circle, fiscal policy, savings, competitiveness, informal econ-omy, sign restrictions, SVAR

JEL code: B23, E26, E62, F32.

∗This paper was prepared for Costin Murgescu Contest (FGDB). It contains 27 pages (including firstpage, bibliography and appendices).†Email: [email protected]

1

1 Introduction

The functioning of fiscal stimulus may adversely affect the normal adjustment mech-anism of the economy, in other words we say we are caught in a vicious circle if anumber of processes are perpetuated in a harmful and reversible way.

The aim of this paper is to investigate a number of three vicious circles that(presumably) act on the Romanian economy and have the power to explain (i) thelack of competitiveness of local firms, (ii) the size of the informal economy and thecatalyst of the three circles, namely (iii) the procyclical behavior of fiscal policy.Croitoru & Tarhoaca (1997) were the first to put forward the mechanism drivingthis disequilibrium by means of the vicious circle of savings, of fiscal policy and ofthe informal economy. With respect to the past period, the economy was still underintense structural reforms and this vicious behavior was to some extent inevitable,but as shown in recent work, Croitoru (2015) reaffirms the presence of the threevicious circles nowadays.

The downside of this papers is the lack of empirical foundation, more precise thepresence of an econometric methodology. A first attempt of modelling this problemwas proposed during my bachelor degree studies: the outcome is that we cannotconfirm the presence of any vicious circle acting on the economy. Unfortunatelysome disadvantages are still present: lack of dynamic interactions, large number ofvariables relative to the small sample. An parsimonious approach is mandatory.

Besides our primary goal, we want to encourage authentic research themes, thattackle domestic problems, since our economy has specific characteristics. Why aren’twe applying general frameworks to the specific problems of our economy and alwaysresort to replicate foreign papers that tell others story? We hope that this paper isonly a starting point.

Here we propose an extended methodology of analysis by using a structural vec-tor autoregressive model (SVAR), estimated using Bayesian techniques and identifythrough sign restrictions approach, derived from the theory of the vicious circles. Allhave the purpose to produce impulse response functions which can be interpretedas structural innovations hitting the economy and can give us economic intuitionabout the presence of the three vicious circles. Using different set of restrictionschemes and variables definition we can obtain robustness of our results and we canmake a reliable analysis for the policy decision side. In this way we are using ageneral framework to analyze the specific problems of our economy.

Before proceeding with this paper it is important to gauge on current economicdevelopment through the means of a set of stylized facts: competitiveness positionof our economy, alongside essential driven factors, the behavior of fiscal policy andtheir impact on external position and the capacity of private and government sectorto adjust.

A comprehensive measure of overall competitiveness and a benchmark to com-pare the performance of economics is the Global Competitiveness Index (GCI),which comprises of 12 pillars, defining the general orientation of an economy (i.e.on factors, efficiency or innovation).

2

According to the classification proposed by Schwab & Sala-i-Martın (2016), Ro-mania is positioned in efficiency-driven bucket of development (the next is innovation-driven economies). As compared to 2006 we can distinguish minor modificationswith respect to each pillar. Infrastructure and the general environment improved(P2, P3), while primary services (P4) actually is lower. Also, a significant improve-ment is in our readiness for technology (P9), but for business sophistication (P11)we have a declining contribution. All in all our competitiveness has improved overthe years and it’s worth mentioning that in 2015 we had the highest score recordedso far and we ranked the 53rd position (see figure 1).

Also within the same report there is available a ranking of the most problematicfactors for doing business: access to financing (generally link to investment deci-sions), inefficient government bureaucracy and tax rates. All of this factors are linkto the conduit of fiscal policy and may be related to the presence of the three viciouscircles.

Figure 1: Competitiveness position

Basic requirements

P1: Institutions

P2: Infrastructure

P3: Macroeconomic environment

P4: Health and primary education

Efficiency enhancers

P5: Higher education and training

P6: Goods market efficiency

P7: Labour market efficiency

P8: Financial marketdevelopment

P9: Technological readiness

P10: Market sSize

Innovation andsophistication factors

P11: Business sophistication

P12: Innovation

P1

P2

P3

P4

P5

P6

P7

P8

P9

P10

P11

P12

2006 2016

06 07 08 09 10 11 12 13 14 15 16

Score 3.9 3.9 4.1 4.1 4.1 4.0 4.0 4.1 4.3 4.3 4.3

7374

68 64

67

7778

76

59

53

62

3.70

3.80

3.90

4.00

4.10

4.20

4.30

4.40

Global Competitiveness Index

Note: Above the bullet is the rank of Romania in each year, while under the axis we have the overallscore of the GCI.Source: Schwab & Sala-i-Martın (2016)

The rise in general competitiveness is linked to several improvements, in whatfollows we present 2 factors driving this upward trend: total factor productivity(whole economy) and real unit labor costs (whole economy)1. Real cost have been

1Defined as ratio of compensation per employee to nominal GDP per person employed — performancerelative to the rest of 37 industrial countries.

3

situated on a declining path starting from the early 2000. Following the abruptcorrection in 2010 (when wages in the public sector were reduced by far as 20%) ourcompensations have been well under European Union benchmark and countries wegenerally compared to2. Total factor productivity (TFP) — indicator of efficiencyand progress — continues to rise on a steady pace, while the other counterpartiesremain on a flat development path (see figure 2). The development of these 2 factorssuggest that our economy has huge potential at disposal: an one hand it runs onefficiency reserves (through lower costs) and on the other it continues to innovate(TFP component).

Figure 2: Factors driving competitiveness

Note: Index: 2010=100; f — forecastSource: AMECO

In order to get a better understanding of the fiscal stance in response to macroe-conomic imbalances we resort to the analysis of fiscal impulse (i.e. negative deltamodifications of structural deficit — a rise translates to a expansionary fiscal policy)vis-a-vis output gap (excess demand from potential level).

In figure 3 we constructed a heatmap with yearly series of fiscal impulse andoutput gap to highlight their comovement. In 70% of the cases the 2 variables hadthe same sign, suggesting the procyclical nature of fiscal policy in response to theeconomic cycle. In 2009 structural deficit was close to 9% of potential GDP, and al-though the efforts was on both sides of the balance sheet (revenue and expenditure),the adjustment process took close to 7 years, over-which revenue and expendituregain similar values 4 to 5 pp (from 31% to 35%, 40% to 35% respectively). Thismeans that the public sector is not flexible on short-run. For 2016 we see a jumpin fiscal stance: a 3% deficit associated to a fiscal impulse of 2.7, while for the same

2Czech Republic, Hungary, Poland.

4

year output gap closed and turn positive (0.7%) and the forecast for 2017-2018strongly suggests the persistence of procyclical behavior.

Figure 3: Fiscal stance

Year Output gap Fiscal impuls

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017f

2018f

Output gap/ fiscal impuls positive

Output gap/ fiscal impuls negative

28

30

32

34

36

38

40

42

44

28

30

32

34

36

38

40

42

44

199

5

199

7

199

9

200

1

200

3

200

5

200

7

200

9

201

1

201

3

201

5

201

7f

Structural deficit Revenue

Expenditure

-8

-6

-4

-2

0

2

4

6

8

199

5

199

7

199

9

200

1

200

3

200

5

200

7

200

9

201

1

201

3

201

5

201

7f

Fiscal impulse Output gap

Note: f — forecastSource: AMECO

The last piece of the puzzle is to add up all the previous results and see howthey affect external position through price-competitiveness indicators and the ad-justment efforts of the 2 sectors (public and private) through their balances. To geta better grip on how the current account (CA) can reflect internal macroeconomicimbalances, it’s worth rewriting it in terms of the twin deficits, starting from basicnational accounts relations3:

Y = C + I +G+NX = C + S + T (A)

S − I︸︷︷︸SNG

+T −G︸︷︷︸SBG

= NX︸︷︷︸CA

(B)

where SNG and SBG stand for private and government balance. Further, inthe scope of this paper, each aggregate level of each balance can be decomposed asthe difference between savings and investment rates:

Sp − Ip︸︷︷︸SNG

+Sg − Ig︸︷︷︸SBG

= ST − IT︸︷︷︸CA

(C)

3With the following legend: Y — income, C — consumption, I — investment, G — governmentexpenditures, NX — net export, S — savings.

5

where:

• Sp, Sg and ST stand for savings in the private and government setor and forthe whole economy.

• Ip, Ig and IT stand for investments in the private and government setor andfor the whole economy.

For macroeconomic stability purpose if CA is -6% and the private sector balanceis -4% and doesn’t have room to adjust anymore, the public budget balance needsto act as a shock adsorbent and keep the deficit at 2%.

Real effective exchange rate (REER) measures the price competitiveness of acountry relative to a set of trading partners. It’s constructed such that a rise isequivalent to a loss in competitive advantage. Developments prior to 2009 werevery volatile, with annual variations of up to 30%, which may be based on the ad-justment process the entire economy was going through. In 2007 the CA plunge tonearly -15% of GDP, with private sector balance (PB) having the largest contribu-tion (around 80%). Remarkably, from a deficit of 11%, the private sector not onlyadjusted, but also managed to be on surplus in less then 2 years. At the same time,the public sector (PB) hit its all time maximum level and the effort to adjust to thetarget of 3% span to 5 years. In the aftermath of the financial crisis, REER wasstable, even gaining competitiveness advantage. With price competitiveness in stillwater, the CA almost closed. Recent developments indicates that firms efforts toadjust their savings are overtaken by the government budget, hence the dive of theCA in 2016 (see figure 4).

Figure 4: Current account development

Source: EUROSTAT

The rest of the paper is structured as follows: Section 2 contains a descriptionof the three vicious circles and a overview of the literature. In the next section,we present the model used, estimation techniques and how they are implemented,followed by results in Chapter 4. Finally, we present final remarks, references anddedicated appendices.

2 Literature Review

The vicious circles in which the Romanian economy is caught were first describedby Croitoru & Tarhoaca (1999); they noticed that when fiscal policy had to reduce

6

its deficit to ensure the macroeconomic equilibrium, decisions were based firstly onthe increase of taxes, and the selection of expenses, depending on necessity andimportance, is bypassed.

Romania’s economy was in a severe recession between 1997 and 1998, and fiscaland monetary policy did not have enough leverage to mitigate the economic down-turn. It should not be neglected that at that time structural changes took placeand were not implemented in a coherent manner, which led to the perpetuation ofthe recession.

Among the structural changes we can mention: the transition to a market econ-omy, the financing of public enterprises was no longer supported by the NBR (es-pecially the agriculture sector), so subsidies were granted by the State to compen-sate for limited credit, consumption was expanding, and the companies could notcope with domestic demand, so imports (both raw material and intermediate prod-ucts) were the main source of trade, so the current account deficit had a massivefall, which required adjustment of the public sector (Croitoru & Tarhoaca (1999),Croitoru (2015)). All these structural changes have put significant pressure on thebudget, and in order to deal with payments, tax increases seemed the fastest solu-tion (not necessarily the most effective, as we will see below).

From the introduction section (equation (B) and (C)) we shown that in the pres-ence of high external imbalances and low adjustment capacity of the private sector,the government needs to step in and make efforts in reducing their deficit. Inthese circumstances, deficit reduction was mandatory for several reasons (Croitoru& Tarhoaca (1999)): (i) the Ricardian equivalence hypothesis is not verified4 (theincrease in the level of saving in the governmental sector is compensated by thesame measure of the reduction of the economies in the non-governmental sector );in line with the reduction of the public deficit, (ii) the real interest rate is reducedwithout influencing the private sector, and (iii) ensuring the macroeconomic balanceimplies the limitation of the current account deficit, as an increase of the govern-ment sector saving will reduce the external imbalance, as the balance of the privatesector remains unchanged.

In what follows we explore the core theory underlying the three vicious circles.Each circles is presented separately.

Vicious circle of savings. In order to meet its macroeconomic stabilizationtarget, the deficit needs to be adjusted in line with the private sector. If this is notpossible, and the funding efforts are focused solely on raising taxes, then the exter-nal imbalance costs are directed to the private sector. Thus, even if they managedto reduce their own deficit, by imposing new taxes, they will be stripped by themand implicitly the necessary resources for investment. In this way, their competi-tiveness can not be improved, which is of the utmost importance, especially whenyou act on a market with strong competition (we can think here of the Euro Area- the main foreign trade partner). Firm’s earnings may not increase (or produc-tion costs are relatively high), and household disposable income is reduced, whichleads to low savings for the formal sector; with limited financial resources, this willlead to an increase in interest rates. With low savings and high interest rates, the

4Christiane & Isabel (2008) show that in the presence of high debt households tend to act as in aRicardian framework.

7

propensity for investment declines. Under these circumstances, efforts to modernizeand increase competitiveness are postponed, and this remains low. As a result, thecircle resumes. The vicious circle of saving is associated with economic downturn,and the fall in the private balance has to be offset by the reduction of the budgetdeficit to the level consistent with the external position sustainability.

Vicious circle of fiscal policy. Fiscal consolidation (reducing the budgetdeficit to levels compatible with external balance) must be done in a rational man-ner: the selectivity of public spending, the financing of public investment projectsthat bring important flows and tax increases in line with fiscal pressure.

The vicious mechanism is triggered by the government’s reduced capacity torationalize spending. Incapacity is favored (in the negative sense) by the majorityof voters that are dependent on redistributions from the state budget. In the givensituation of the private sector deficit, balance stabilization is achieved by increasingthe tax rates and random expenditure cuts. Negative effects on savings are spreadover 2 channels: (i) public resources will continue to scatter on unsustainable socialsecurity schemes and social assistance programs, along with the promotion of inef-ficient public-funded projects5 and (ii) tax increases affect the economic position offirms, especially when the ability to adjust to new taxes is limited. As an economiccontraction can be induced, a new deficit adjustment is necessary, and as long asvoters dependent on redistributions are overwhelming, the government cannot ra-tionalize public spending. This in turn trains the only way of adjusting the deficit,i.e. raising taxes, so corruption and uncertainty of property rights are maintainedand the vicious circle of saving is fed.

Vicious circle of the informal economy. The other negative influence result-ing from the reduced capacity of the government to select expenditures is triggeringthe vicious circle of the informal economy. As tax pressure rises, as a result of taxincreases, companies decide to migrate towards informal sector. The expansion ofthe informal sector leads to a reduction in the tax base and, apparently, the level ofcollection is reduced. With a reduced base and the need for fiscal stabilization, newtax increases are required, thus increasing the transfer of firms to the informal sec-tor. Davis & Henrekson (2004) confirms that there is a positive relationship betweenthe level of taxation and the informal economy, especially in developed countries,with high income levels. Chiarini, Marzano & Schneider (2013) talk about a viciouscircle between fiscal pressure and evasion, but its presence in Italy is not confirmed.

In figure 5 we present the 3 vicious circles that describe all the above mentioninteractions through a series of flow charts.

A first attempt of modelling this problem was proposed during my bachelordegree studies: using a cointegration methodology, I’ve selected a set of proxy vari-ables describing the dynamics of the circle and I’ve estimated a long run and a errorcorrection mechanism, individually. The validation criteria was based of the sign ofthe estimated coefficients, their speed of adjustment to the long run relation, thelink to a common variable (taxation rate) and also impulse response functions.

5Public projects have had a central role to play in managing investments, precisely to support struc-tural reform, but political guidelines generally have the last word.

8

Figure 5: Interactions between the 3 vicious circles

Lowcompet-itiveness

Vicious Circleof Savings

Lowsavings

High realinterest

rates

Investmentsare dis-

couraged

Firm’sefforts arepostponed

Publicexpenses

are not ra-tionalized

Vicious Circleof Fiscal Policy

Budgetdeficit

is rising

Ad-hoctax in-

creases orineffcientexpendi-tures cut

New ad-justment isnecessary

Dependentvoters onredistri-bution

Taxesincrease

Vicious Circle of theInformal Economy

Fiscalburden

is higher

Firmsmigrate toinformalsector

Tax baseis lower

Collectionrate

decreases

Note: solid line - direct effect (within the circle), dotted line - reiteration of the circle, red line - indirecteffect (outside the circle) which exacerbates the propagation mechanism.Source: personal interpretation based on Croitoru & Tarhoaca (1999), Croitoru (2014).

9

The outcome is that in a weak sense (not taking into account the interlinkagesbetween the 3 circles), only the vicious circle of savings and the vicious circle of theinformal economy is present. On the other hand, in a strict manner, the one im-posed by the authors, we cannot confirm the presence of any vicious circle acting onthe economy. Unfortunately some disadvantages are still present: lack of dynamicinteractions between the three equations and large number of variables relative tothe small sample (for more details see Stavre (2015)).

As mention in the introduction section we impose sign restrictions based on thetheory of the authors. This approach is widely used in the literature: Peersman &Straub (2004) and Peersman & Straub (2006) analyze the effects of a technologyshock on the labor market, using sign restrictions derived from the Real BusinessCycle model and New Keynesian model; other examples are Uhlig (2005) for theimpact of monetary policy shock on output and Mountford & Uhlig (2005) for theeffects of a fiscal shock.

3 Methodology

In this section we present the theoretical model together with estimation techniques,identification schemes adopted, as well as the data used.

To meet our objective we apply the following methodology: (i) select proxy vari-ables that summaries components of the vicious circles, (ii) estimate VAR model,(iii) derive sign restrictions from the theory of the authors, (iv) make structuralinferences based on SVAR and (v) analyze the response of the unrestricted vari-ables. The validation process is straightforward: check whether the response ofthe unrestricted variables match the ones implied by the theory of the authors. Indoing so, we analyze IRF in two ways: (i) plot the dynamic response over time,along side the confidence interval to make a decision about the sign and (ii) ana-lyze the full distribution of results for 4 quarters by comparing the median with zero.

We begin with the setup of the model. Let Yt = (y1,t · · · yn,t)′ be a vector ofendogenous variables and consider the standard VAR(p) model representation:

Yt = c+A1Yt−1 + · · ·+ApYt−p + ut, (1)

where ut is a n-dimensional vector of Gaussian residuals (ut ∼ N(0,Ψ) andEutu

′t = Ψ), c = (c1, · · · , cn)′ is a n-dimensional vector of intercepts and A1, · · · , Ap

are n× n autoregressive matrices.

This model can be estimated using classic frequentist techniques via OrdinaryLeast Squares (OLS) estimator. It is convenient to re-write the model in equation(1) as a system of multivariate regressions:

Y︸︷︷︸T×n

= X︸︷︷︸T×k

B︸︷︷︸k×n

+ U︸︷︷︸T×n

, (2)

where Y = (Y1, · · · , YT )′, X = (X1, · · · , XT )′ with Xt = (Y ′t−1, · · · , Y ′t−p, 1),U = (u1, · · · , uT )′ and B = (A1, · · · , Ap, c)

′ is the k × n matrix containing all

10

coefficients and k = np + 1. OLS estimator and the computed estimate for thecovariance matrix are given by:

B = (X ′X)−1XY

Ψ =1

T − k − 1(U ′U)

(3)

The most important shortcoming of OLS estimator when applied to VAR modelsis dimensionality: in large systems the estimator becomes bias. To tackle thisproblem we resort to Bayesian estimation. In this alternative approach our priorbelieves can be incorporated in estimation and standard Bayesian VAR (BVAR)techniques builds on the assumption that all the equations are ”centered” around arandom walk with drift6 and that coefficients can be shrink, this is what Litterman(1985) refers to as Minnesota prior and the generalization is given by a NormalInverted Wishart prior:

vec(B)|Ψ ∼ N(vec(B0),Ψ⊗ Ω0) and Ψ ∼ iW (S0, α0) (4)

where B0,Ω0, S0 and α0 are chosen so that prior believes and variances of B andexpectations of Ψ are equal to those implied by the Minnesota prior.

The disadvantages of using Minnesota prior is that we cannot estimate largeBVAR models, no prior covariance is assumed among the VAR coefficients and withthis kind of structure is very difficult to work, especially when dealing with unitroot processes7. To tackle with this shortcomings, in what follows we are goingto impose the previous prior by means of dummy observations construction as inBanbura, Giannone & Reichlin (2008). We start by adding Td dummy observationsYd and Xd to the system in equation (2) and impose the Normal Inverted Wishartprior with:

B0 = (XdX′d)X ′dYd Ω0 = (XdX

′d)−1 S0 = (Yd −XdB0)

′(Yd −XdB0) α0 = Td − k

For matching Minnesota moments, the dummy variables have the following struc-ture:

Yd =

diag(δ1σ1, · · · , δnσn)/λ0n(p−1)×n· · ·

diag(σ1, · · · , σn)· · ·

01×n

Xd =

Jp ⊗ diag(σ1, · · · , σn)/λ 0np×1

· · · · · ·0n×np 0n×1· · · · · ·

01×np ε

(5)

where Jp = diag(1, 2, · · · , p), δ1, · · · , δn are prior means for each coefficients, λis a hyperparameter that control for the overall tightness of the prior around therandom walk process8 and σ1, · · · , σn are scaling parameters and are approximatedwith the variance of the univariate process of each variable in the VAR model. Ydimposes prior beliefs on the autoregressive coefficients, the first block of Xd arepriors for the covariance matrix, while the second block is an uninformative prior

6More precisely an AR(1) process: Yt = c+ Yt−1 + ut.7For a comprehensive survey on the literature of prior used for BVARs see Dieppe, Legrand & Roye

(2016).8Set λ = 0 then the posterior equals the prior and data has no influence on the estimates and for

λ =∞ posterior coincides with OLS estimates.

11

for the intercept (if ε is set to very small values). Augmenting the model in equation(2) with the dummies in equation (5) we get:

Y∗︸︷︷︸T∗×n

= X∗︸︷︷︸T∗×k

B︸︷︷︸k×n

+ U∗︸︷︷︸T∗×n

, (6)

with the following mapping:

T∗ = T + Td, Y∗ = (Y ′, Y ′d)′ X∗ = (X ′, X ′d) U∗ = (U ′, U ′d)

and the final posterior is given by:

vec(B)|Ψ, Y ∼ N(vec(B,Ψ⊗ (X ′∗X∗)−1) and Ψ|Y ∼ iW (Σ, Td + 2 + T − k)

(7)

with B = (X ′∗X∗)−1X ′∗Y∗ and Σ = (Y∗ − X∗B)′(Y∗ − X∗B). The last step is to

resort to Gibbs sampler in order to get the estimated values.

In order to gauge on the effects of the three vicious circles on the Romanianeconomy we need to identify this shock through an SVAR model. Consider thereduced form representation in equation (1) and re-write in AR(1) representation:

Yt = B · Yt−1 + ut with ut ∼ N(0,Σu) (8)

In equation (8) the residuals are correlated, which reflect contemporaneous relationbetween variables. In order to disentangle shocks by economic meaning we needfor the innovations to be uncorrelated. With this in mind we need a structuralrepresentation of our economy:

A · Yt = B · Yt−1 + et with et ∼ N(0, I) (9)

Matrix A contain all contemporaneous effects. We cannot estimate directly thismodel, but if we multiply equation (9) with A−1

A−1 ·A · Yt = A−1 ·B · Yt−1 + A−1 · et ⇒ Yt = F · Yt−1 + ut (10)

we get a reduced form, which we can now estimate. Combining this observationwith equation (8) and (9) we have:

F = A−1 ·But = A−1 · et

Σu = A−1 · I ·A−1′ = A−1 ·A−1′(11)

Identification is achieved is we can pin down matrix A−1. The most basic identi-fication scheme is to assume a recursive chain, hence Choleski factorization allowsidentification, if P is a lower triangular matrix since:

Σu = A−1 ·A−1′ = P ′ · P (12)

After identification is achieved we analyze the impulse response function (IRF):how a variable would respond if a structural shock is applied. Previous restric-tions are know as ”short run” restrictions. An alternative approach is to use signrestrictions on the response of variables to various shocks, derived from economictheory (e.g. a monetary policy shocks must have a negative impact on inflation).

12

For details see Uhlig (2005). A problem with Choleski factorization is that the de-composition in equation (12) is not unique since we can use any orthogonal matrixS (with the property that S′S = I) and write:

Σu = P ′ · P = P ′ · I · P = P ′ · S′ · S · P = R′ ·R (13)

This new matrix is no longer lower triangular, but the information contain inΣu is not altered, in this case, depending on the numbers of random draws, we getthe same number of possible decompositions of the covariance matrix. To be ensurethat our random matrix is good we verified it against a set of sign restrictions witheconomic meaning (imposed a priori). To obtain the IRFs we must proceed withthe next algorithm (Blake & Mumtaz (2012)):

1. Estimate the reduced-form VAR and obtain F and Σu

2. Compute P ′ = chol(Σu).

3. Draw a random orthogonal matrix S.

4. Compute A−1 = R = P ′S′.

5. Compute the impulse response associated with R(= A−1).

6. Are the sign restrictions satisfied?

(a) Yes ⇒ store the impulse response.

(b) No ⇒ discard the impulse response.

7. Repeat 3-6 until we obtain N replications.

8. Report mean or median and confidence interval (in general 16th and 84thpercentiles).

Proxy variables are required for the model. For the vicious circle of savingswe need a measure for competitiveness, the most ovious one is to use real effectiveexchange rate (REER)9. We combine savings and investment in the form of pri-vate sector balance (SNG) and use real interest rate for loans contracted by firms(RFIN). For the next vicious circle (fiscal policy) we need a variable that capturesthe majority of voters that are dependent on redistributions; we decided to make useof capital expenditures and total transfers. These two variables show the orientationof the budget (towards productive expenses or inefficient schemes); to summarizethem we make a ratio of them and they denote a discretionary measure (DISCR).A proxy variable to capture the ad-hoc tax increases (i.e.: tax burden) is implicittaxation rate (ITR). This measure is used to compare different effective averagetax burden levied on different types of income sources. Alongside the previous 2variables we include the government budget balance (SBG).

The last challenge arises from the approximation of the vicious circle of the in-formal economy. For this we calculate a measure of informal economy sector (IES).Davidescu & Dell’Anno (2016) provide a comprehensive set of alternative measureconducted on the Romanian market (MIMIC approach, labor approach etc.). Thereresults are fairly robust to different specifications and with this in mind we chooseto implement the simpler approach, namely the currency demand model, since wedon’t want to draw attention from the main objective 10. We have 2 constructed

9REER based on labor cost or prices may tend to show conflicting signals, but both are relevant forthe dynamics of export (Claire & Francesco (2015)). They also suggest using a non-price factors (e.g.product quality, industry specialization, efficiency of sales networks etc) in the form of absolute andrelative total factor productivity, constructed in a similar manner as REER.

10We must keep in mind that our primary concern are the 3 vicious circles.

13

variables (i.e.: level of taxation and size of informal economy). In order to keep themethodology as simple as possible we present the details in dedicated appendices.In appendix B and C we present the complete process used to obtain overall ITRand IES.

The final step is to derive sign restrictions used for the identification schemes.From diagram 5 we can obtain the set of restrictions for our variables. We do nothave in hand a theoretical model, but we have the expected sign of the variablesand can used them as prior information to pin down the structural shocks. As aconsequence, each of this shocks are interpreted as shocks triggering the appearanceof the vicious circles. In table 1, 2 and 3 we highlight the restrictions used. Forscheme S1 we have a basic set of restrictions derived from the authors theory, mov-ing to scheme S2 we add restrictions for the shock triggering the vicious circle ofthe informal economy and lastly add more restrictions for every shock, in order tobetter identify them. Also, when implementing the sign restrictions we are agnosticwith respect to time horizon (i.e. we restrict the sign for 0 periods, thus data canfreely inform the variables response).

After the model is identified we are going to analyze the response of the un-restricted variables and compared it with the theoretical expected sign (e.g. forthe shock triggering the vicious circle of savings, for identification scheme S1, weare going to look at all the variables except SNG, REER, RFIN; next the othershock and so on and so forth). We are aware that we do not have an benchmarkstudy to compare our results (under an empirical approach). For that we resort torobustness analysis, to ensure that our results remain a fact: use alternative identi-fications schemes and specify REER under 2 alternative deflators: consumer priceindex (CPI) and unit labor cost (ULC).

Table 1: Sign restrictions (scheme S1)

SBG SNG REER RFIN DISCR IES ITR

VC of savings ↓ ↑ ↑VC of fiscal policy ↓ ↓ ↑VC of the informal economy ↑



VC of savings ↓ ↑ ↑VC of fiscal policy ↓ ↓ ↑VC of the informal economy ↓ ↑ ↑



VC of savings ↓ ↑ ↑ ↑VC of fiscal policy ↓ ↑ ↓ ↑VC of the informal economy ↓ ↑ ↑ ↑

Note: VC — shock triggering vicious circle.

14

Data spans the period 2000Q1-2016Q4, we include 4 lags for the VAR model,according to information criteria. Data source is National Bank of Romania (NBR),National Institute of Statistics (NIS) and EUROSTAT. We stop with the simula-tions at 50000 replications due to the computation burden11. All computations weredone in MATLAB R© 2017a12.

The main tools of our empirical assessment are Impulse Response Functions(IRF), Forecast Error Variance Decomposition (FEVD) and Historical Decomposi-tion (HD).

4 Results

In this section we present the main findings of our empirical work: IRFs, FEVDsand HDs, including robustness switches. Due to page limitation we present only therelevant results. We start by comparing the response of the unrestricted variablesto the shocks triggering the vicious circles. In what follows we analyze the modelthat includes the measure of competitiveness (i.e. RERR) based on the CPI deflatorand is identify using scheme S1.

For the shock triggering the vicious circle of savings we get the expected sign fordiscretionary measure and for informal economy sector, while the other 2 remainingdon’t validate: the response of government budget is positive and significant foralmost 10 periods, while ITR has a huge hike in the first 5 quarters; suggestingmaybe the implementation lag that fiscal policy usually encounters.

From the most unrestricted shock of this scheme (the one fueling the appear-ance of informal economy) result are even: 3 variables have the expected sign (publicbudget, REER, discretionary variable) and the remaining do not validate; also nonehas statistical relevance and for REER and ITR the adjustment process (return tozero) is very slow.

The shock that is intended to be the catalyst of the three vicious circle, namelythe shock driving the vicious behavior of fiscal policy has only one sign that confirmsthe theory of the authors (the rise of real interest rate). For private sector balancewe have a relevant and positive response and informal economy in reduced by ahuge portion (compared to the later adjustment). Real effective exchange rateinitially benefits from the triggering of the vicious circle of fiscal policy, but withina year suffers a huge loss in price competitiveness (this may describe the short termbenefits of a fiscal policy). All results are presented in figure 6. When movingto more restrictive identifications schemes (S2 and S3), remarks tend to remainconstant: the unrestricted variables preserve their sign when switching to otherschemes. This consistency suggest that our results are robust. Complete set ofresults are presented in appendix D (figure 12 and 13).

11Given that we have to run the algorithm six times (3 alternative identification schemes and 2 REERmeasures), the total elapsed time was close to 5 days.

12We build on the scripts available on Haroon Muntaz’s and Ambrogio Cesa-Bianchi’s personal website.

15

Figure 6: Response of unrestricted variables (identification scheme S1)

(a) Shock triggering the vicious circle of savingsGovernment budget balance

0 5 10 15 20 25 30 35

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

10-3 Discretionary measure

0 5 10 15 20 25 30 35

-10

-8

-6

-4

-2

0

2

4

6

810-3

Informal economy sector

0 5 10 15 20 25 30 35

-2

0

2

4

6

8

10

10-3 Implicit taxation rate

0 5 10 15 20 25 30 35-1.5

-1

-0.5

0

0.5

1

1.5

10-3

(b) Shock triggering the vicious circle of fiscal policyPrivat sector balance

0 5 10 15 20 25 30 35

-1

0

1

2

3

4

5

10-3 Real effective exchange rate

0 5 10 15 20 25 30 35

-4

-2

0

2

4

6

8

10-3

Real interest rate

0 5 10 15 20 25 30 35

-1

0

1

2

3

4

10-3 Informal economy sector

0 5 10 15 20 25 30 35

-12

-10

-8

-6

-4

-2

0

2

4

10-3

(c) Shock triggering the vicious circle of the informal economyGovernment budget balance

0 5 10 15 20 25 30 35

-5

0

5

10

1510-4 Privat sector balance

0 5 10 15 20 25 30 35

-3

-2.5

-2

-1.5

-1

-0.5

0

0.5


0 5 10 15 20 25 30 35

-6

-5

-4

-3

-2

-1

0

1

10-3

Real interest rate

0 5 10 15 20 25 30 35

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2


0 5 10 15 20 25 30 35-5

0

5

10


0 5 10 15 20 25 30 35

-1

-0.5

0

0.5

1

1.5

10-3

Source: own computations.

16

As an additional measure to verify the correctness of the responses of variablesto the structural shocks, we also analyze detailed distribution of each variables, overa horizon of 4 quarters. In figure 7 we plot the entire distribution of results, themedian and zero benchmark line. A response is valid if the median is situated over/below the zero line and respect the theoretic sign. We present the results only foridentification scheme S3 (since the other 2 where similar).

Figure 7: Distribution of IRFs (identification scheme S3)

Note: Red line is the zero benchmark; blue line is the median; VC — shock triggering vicious circle; Tstands for period (e.g. T=1 is the first quarter after the shock took place).Source: own computations.

The assessment presented above remain unchanged. As an overview across dif-ferent schemes and definition for variables, we get that 4 out of 10 unrestrictedvariables have the expected sign derived from the theory of the authors. In theseconditions we cannot point with the conclusions in the same directions as the au-thors, hence the three vicious circles are not present in our economy (at least asspecify in their paper).

In table 4, a forecast error decomposition exercise is presented — each responseof the variable, over a specific horizon, is explain by certain contribution of a specificshock; all structural shocks sum up to 100%. Figures are based on identificationscheme S3 (the other 2 show similar patterns) over 5, 10 and 40 quarters horizon.

17

Table 4: FEVD — identification scheme S3

Variable Shocks 5Q 10Q 40Q

SBG

Savings 13.9% 13.7% 15.1%Fiscal 11.1% 11.1% 12.4%

Informal 15.0% 15.1% 16.0%Others 60.1% 60.1% 56.5%

SNG

Savings 17.2% 18.7% 18.7%Fiscal 11.2% 11.4% 12.8%

Informal 22.0% 21.7% 19.6%Others 49.5% 48.2% 48.9%

REER

Savings 19.4% 21.7% 19.5%Fiscal 14.6% 17.6% 20.4%

Informal 13.5% 13.4% 13.3%Others 52.5% 47.3% 46.8%

RFIN

Savings 10.6% 11.1% 12.7%Fiscal 13.2% 13.6% 14.3%

Informal 13.7% 13.8% 14.6%Others 62.4% 61.5% 58.4%

DISCR

Savings 13.3% 13.4% 14.8%Fiscal 13.8% 14.5% 15.0%

Informal 13.8% 13.7% 14.6%Others 59.1% 58.4% 55.6%

IES

Savings 15.3% 16.0% 16.6%Fiscal 12.9% 12.8% 13.8%

Informal 20.8% 21.1% 19.8%Others 50.9% 50.1% 49.8%

ITR

Savings 12.0% 12.2% 13.0%Fiscal 10.8% 12.3% 16.1%

Informal 14.1% 13.8% 13.4%Others 63.1% 61.7% 57.5%

Note: Others refer to the cumulative value of the remaining 4 unidentified shocks.

When adding the three shocks we get o combine marginal contribution of 43%for 5 quarters horizon, while for the entire horizon of 10 years it slightly increasesto 47%. The contributions of shocks is roughly equal across different identificationsscheme: for the shock triggering the vicious circle of savings the impact range is14.6% - 15.8%, the next shock (fiscal) the values are 12.5% - 15% and for the lastshock considered the interval is wider, but also decreasing when moving to distanthorizons (16.1% - 15.9%). While is impact dissipates over time, the shock triggeringthe appearence of the vicious circle of informal economy remains the most important.

The last tool in our empirical analysis is the historical decomposition exercise:each variable is presented as deviation from baseline and is decompose in structuralshocks derived from the SVAR model. In figure 8 we plot only the three shockstriggering the appearance of the vicious circles.

The shock triggering the vicious circle of savings had the most prominent impacton the real effective exchange rate (alongside the shock responsible for the informaleconomy) and on the government budget balance (especially in 2009). The size ofthe informal economy was in second stage affected by the fiscal policy conduit (theown shock has the higher effect). Real interest rate was mainly driven by the shocktriggering the informal economy circle.

18

Figure 8: Historical decomposition (identification scheme S3)

Government budget balance

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1-0.08

-0.06

-0.04

-0.02

0

0.02

0.04Privat sector balance

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1-0.1

-0.05

0

0.05

0.1Real effective exchange rate

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1-0.1

-0.05

0

0.05

0.1

0.15

0.2

Real interest rate

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1-0.1

-0.05

0

0.05

0.1Discretionary measure

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1-0.2

-0.1

0

0.1

0.2

0.3

0.4Informal economy sector

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1-0.2

-0.1

0

0.1

0.2

Implicit taxation rate

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1-0.03

-0.02

-0.01

0

0.01

0.02

0.03

VC of SavingsVC of Fiscal policyVC of Informal economy

Note: VC — shock triggering vicious circle; results are presented as deviation from baseline.Source: own computations.

Also, as emphasized in the previous remarks, the most important structural shockobtain from the FEVD exercise (informal economy) is present in all variables withrelative important contributions. There are also some variables that have beenequally affected by the three shocks, although in small size, but the remaining 4unidentified shocks explain most of the recorded deviations.

Our last comment is with respect to the robustness analysis: when switchingto the real effective exchange rate measured by the unit labor cost deflator results(IRF, FEVD or HD, across all 3 identification schemes) do not register significantchanges, thus we only mention them here13.

13Complete set of results are available upon request.

19

5 Final remarks

Throughout this paper we intended to develop a validation process for testing thethree vicious circles that (presumably) act on the Romanian economy, as describedby Croitoru & Tarhoaca (1997) and Croitoru (2015). We build an SVAR modelusing proxy variables to capture the interactions between the circles and imposesign restrictions derived from the aforementioned papers and analyze the responseof the variables using different identifications schemes and definitions of variables.

All in all, our results do not harmonized with the theory of the authors. As anoverview we get that 4 out of 10 unrestricted variables have the expected sign. Thuswe cannot confirm the presence of these circles in our economy (at least with thecurrent methodology — a dose of conservatism must be maintained when treatingthis finding as a final result).

The competitiveness of the Romanian economy has improve over time, real laborcosts and total factor productivity are supportive for future growth. Nowadays, thecurrent account is manageable, but recent developments indicates that the publicbudget has started a new process of taking away the savings that the private sectorhas strive to conserve.

Policy implications are pretty straight forward. Although the three circles havenot been confirmed, the recommendation for policy makers is valid: awareness of avicious circle acting on the economy is the first step in eliminating the componentthat generates perturbation in the transmission mechanism.

We also have to take into account future redevelopment and better ways toimprove the underlying paper, in this respect we propose to: (i) search for newvariables, (ii) build a structural model and (iii) use an optimal rate to gauge thepoint at which the tax burden induces disequilibrium for firms.

Lastly we have faith that this paper will encourage future research with domesticconsiderations as the main driver.

20

6 References

Banbura, M, Giannone, D & Reichlin, L 2008, ’Large Bayesian VARs’, ECB Work-ing Paper Series, no. 966.

Blake, A & Mumtaz, H 2012, ’Applied Bayesian econometrics for central bankers’,CCBS Technical Handbook No. 4.

Cagan, Philip, 1958. The Demand for Currency Relative to Total Money Supply.The Journal of Political Economy, 66, 303–29.

Chiarini, B, Marzano, E & Schneider, F 2013, ’Tax rates and tax evasion: anempirical analysis of the long-run aspects in Italy’, European Journal of Law andEconomics, vol 35, no. 2, pp. 273-293.

Christiane, N & Isabel, V 2008, ’Fiscal policies, the current account and ricardianequivalence’, ECB Working Papers, no. 935.

Claire, G & Francesco, Z 2015, ’Exploring price and non-price determinants oftrade flows in the largest euro-area countries’, ECB Working Papers, no. 1789.

Croitoru, L & Tarhoaca, C 1999, ’Fiscal Policy in Romania’, RCEP WorkingPapers, no. 14.

Croitoru, L 2015, ’Cercul Vicios al Politicii Fiscale’, Studii BNR.

Davis, SJ & Henrekson, M 2004, ’Tax Effects on Work Activity, Industry Mixand Shadow Economy Size: Evidence from Rich-Country Comparisons’, NBERWorking Papers, no. 10509.

Davidescu, A & Dell’Anno, R 2016, ’What is the size of the shadow economy inromania? A comparison of different estimation approaches’, SIEP 2016.

Dieppe, A, Legrand, R & Roye, BV 2016, ’The BEAR toolbox’, ECB WorkingPapers Series, no. 1934.

Litterman, RB 1985, ’Forecasting with Bayesian Vector Autoregressions - FiveYears of Experience’, Federal Reserve Bank of Minneapolis Working Paper, no.274.

Mountford, A. & H. Uhlig (2005), “What are the Effects of Fiscal Policy Shocks”,Discussion Paper 2005-039, SFB 649, Humboldt-Universitat, Berlin.

Peersman, G & Straub, R 2004, ’Technology Shocks and Robust Sign Restrictionsin a Euro Area SVAR’, ECB Working Papers, no. 373.

Peersman, G & Straub, R 2006, ’Putting the New Keynesian Model to a Test’,IMF Working Papers, no. 135.

Schwab, K & Sala-i-Martın, X 2016, ’The Global Competitiveness Report’, WorldEconomic Forum, Geneva.

Stavre, IG 2015, ’Testarea celor 3 cercuri vicioase ale economiei romanes,ti: oabordare VAR–VEC’, Colect, ia de working papers ABC-ul Lumii Financiare, no.3.

Tanzi, Vito, 1983. The Underground Economy in the United States: AnnualEstimates, 1930-80. International Monetary Fund Staff Papers 30, 283–305.

Taxation Trends, 2017, ’Taxation Trends in the European Union (Technical ap-pendix of the document, section F)’.

Uhlig, H 2005, ’What are the effects of monetary olicy on output? Result froman agnostic identification procedure’, Journal of Monetary Economics, no. 52, pp.381-419.

21

A Data

In this section we described and present the data used in estimation.

Government budget balance. Difference between government total revenueand government total expenditure (normalized over GDP). Source: EUROSTAT.

Privat sector balance. The variable is constructed by substracting govern-ment budget balance from current account (normalized over GDP). Source: EURO-STAT.

Real effective exchange rate. Deflator: consumer price indices (CPI) andunit labour costs in the total economy (ULC), both for 37 trading partners (index:2005=100). A rise is equivalent to a loss of competitiveness. Source: EUROSTAT.

Real interest rate. Interest rate for loans contracted by non-financial corpora-tions (outstanding), adjusted for inflation (based on GDP deflator). Source: NBR,NIS.

Discretionary measure. The variable is constructed as a ratio between capitalexpenditures (gross capital formation) and total transfers (social benefits, capitaltransfers and other transfers — all payable) of the government. Source: EURO-STAT.

Informal economy sector. Constructed variable — see appendix C. Source:NBR, NIS, EUROSTAT.

Implicit taxation rate. Constructed variable — see appendix B. Source: EU-ROSTAT.

Figure 9: Variables used in estimation

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1

-0.1

-0.08

-0.06

-0.04

-0.02

0

Government budget balance

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1

-0.1

-0.08

-0.06

-0.04

-0.02

0

0.02

0.04

0.06

Privat sector balance

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1

4.3

4.4

4.5

4.6

4.7

4.8

4.9Real effective exchange rate (ULC)

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q14.4

4.45

4.5

4.55

4.6

4.65

4.7

4.75

Real effective exchange rate (CPI)

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1

-0.02

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

Real interest rate

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0.55

0.6

0.65Discretionary measure

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0.55Informal economy sector

2000Q1 2003Q1 2006Q1 2009Q1 2012Q1 2015Q1

0.21

0.22

0.23

0.24

0.25

0.26

0.27

Implicit taxation rate

Source: NBR, NIS, EUROSTAT, own computation.

22

B Implicit Taxation Rate

In this section we described the process used to obtain Implicit Taxation Rate (ITR),using the methodology described in Taxation Trends (2017).

The general concept of ITR is to identify all eligible taxes for an category (e.g.consumption) and a suitable denominator (i.e. potential tax base), such that theratio between the aforementioned series produces the desire estimation of the taxrate. Main data source is the information from national accounts14. After comput-ing the 3 individuals rates (for consumption — C, labor — L and capital — K),overall ITR is constructed as a weighted average, where weights are obtain from therelative contribution of individual taxes to total revenue of the government.

We aggregated and compared our results with the annual values provided byTaxation Trends report (only for consumption and labor). For ITRC we have lowervalues (with 4.91 pp), while for ITRL higher (3.45 pp). Although our figures areslightly different, when calculating the correlation coefficients for both componentwe have over 90%. Taking this into account, we have a very reasonable proxy fortaxation and can compute the overall ITR (see figure 10).

Figure 10: Results for Implicit Taxation Rate

0

5

10

15

20

25

30

20

00

20

02

20

04

20

06

20

08

20

10

20

12

20

14

20

16

Decomposition of ITR total

C L K

0.96

0.91

0

10

20

30

40

200

32

00

42

00

52

00

62

00

72

00

82

00

92

01

02

01

12

01

22

01

32

01

42

01

5 C L

Qualitative check

C (Taxation Trends)L (Taxation Trends)C (Calculated)L (Calculated)Correlation coefficient

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

19

99Q

1

20

01Q

3

20

04Q

1

20

06Q

3

20

09Q

1

20

11Q

3

20

14Q

1

20

16Q

3

Implicit Taxation Rates

C L K ITR

Note: ITR stands for overall ITR, while C/ L/ K are the subcomponents defining consumption/ labor/capital.Source: EUROSTAT, Taxation Trends (2017); own computations.

14The main challenge is that ITRs are computed using a more refined database at annual frequency,but our current need is for quarterly frequency.

23

C Informal Economy Sector

In this section we described and present the derivation of the informal economy size,using the currency demand model. This type of monetary model builds on the workof Cagan (1958) and Tanzi (1983). In short, demand for currency is dependent on aset of factors: economic activity, taxation level, interest rate; the first 2 are expectedto have a positive impact, while the last variables captures a opportunity cost andthe expected sign is negative. Once we have a measure for level of currency, we canignore the illegal factor (i.e. taxation) and thus obtain the supplement of currency.

Following Davidescu & Del Negro (2016) we estimate a Vector Error CorrectionModel (VECM) using the following variables: real currency (nominal currency incirculation deflated with the GDP deflator) — CR, real GDP — Y , inflation (HICPbased) — P , tax revenues (over GDP) — T , nominal interest rate (term deposits)— I and employment rate — E. All variables are specified as ln(1+x), with the ex-ception of currency and real GDP who are tranformed as simple ln(x). Estimationperiod is 2000-2016 (quarterly frequency), source of data: NIS, NBR, EUROSTAT.Information criteria indicates the usage of 3 lags and Johansen cointegration testsuggest at least one cointegration vector.

Once parameters are obtain through OLS estimation we get an estimated valuefor currency (CR), next the tax variable (Tt) and inflation (Pt) are set to minimumlevel15 and the model is solved to extract the level of currency without tax burden(CRNT ). Subtract the previous 2 variables and the additional currency reflects thepart generated by underground economy. The last step to obtain informal GDP(Yi) is to make a ratio of shadow currency (Ci) and official currency (Cf ), correctthe estimates for the fact that income elasticity (i.e. β = 1.17) is not unitary andto multiply with official GDP level (Yf )16. The final value obtained is normalizedover GDP.

In table 5 and figure 11 we present the results for the long-run equation fromthe VECM estimation. Model has good statistical properties: with exception ofinterest rate, all estimated parameters have the expected sign and are significant.Fitted values are within the confidence interval and residuals are not correlated orheteroskedastic.

From the historical decomposition exercise, demand for currency experience asudden rise in 2007 due to a higher tax development and in recent years was mainlydriven by rise in employment and economic activity, while in 2016 we distinguishnegative contribution from taxes.

15The taxes over GDP ratio are fixed at 30%, which translates to the value of 0.2624 and inflation ata conservative 0%.

16Defined as Yi = ( Ci

Cf)β

−1 · Yf .

24

Table 5: Long run equation results

Parameter Value

Y 1.17(***)

P11.64

( )

T10.75(***)

I-3.04(***)

E11.56(***)

Intercept -11.65(***)

StatisticsR2 adj 46%LM test 0.91White test 0.35

Note: ***, **, * denote statistically significant at 1%, 5% and 10%.H0 (LM test): no serial correlation at lag order 5 (similar results for inferior lags).H0 (White test): no heteroskedasticity.Source: own computations.

Figure 11: Informal economy estimation results

Source: NIS, EUROSTAT; own computations.

25

D Detailed results



0 5 10 15 20 25 30 35

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5


0 5 10 15 20 25 30 35

-10

-8

-6

-4

-2

0

2

4

6

810-3

Informal economy sector

0 5 10 15 20 25 30 35

-2

0

2

4

6

8

10


0 5 10 15 20 25 30 35-1.5

-1

-0.5

0

0.5

1

1.5

10-3


0 5 10 15 20 25 30 35

-1

0

1

2

3

4

5


0 5 10 15 20 25 30 35

-4

-2

0

2

4

6

8

10-3

Real interest rate

0 5 10 15 20 25 30 35

-1

0

1

2

3

4


0 5 10 15 20 25 30 35

-12

-10

-8

-6

-4

-2

0

2

4

10-3


0 5 10 15 20 25 30 35

-5

0

5

10


0 5 10 15 20 25 30 35

-6

-5

-4

-3

-2

-1

0

1

10-3

Real interest rate

0 5 10 15 20 25 30 35

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2


0 5 10 15 20 25 30 35-5

0

5

10

10-3


26



0 5 10 15 20 25 30 35

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5


0 5 10 15 20 25 30 35

-10

-8

-6

-4

-2

0

2

4

6

8


0 5 10 15 20 25 30 35

-2

0

2

4

6

8

10

10-3


0 5 10 15 20 25 30 35

-1

0

1

2

3

4

5

10-3 Real interest rate

0 5 10 15 20 25 30 35

-1

0

1

2

3

4


0 5 10 15 20 25 30 35

-12

-10

-8

-6

-4

-2

0

2

4

10-3


0 5 10 15 20 25 30 35

-5

0

5

10

15

10-4 Real interest rate

0 5 10 15 20 25 30 35

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2


0 5 10 15 20 25 30 35

-5

0

5

10

10-3


27

testing the three vicious circles of the romanian economy: a … · 2017-12-11 · testing the...

Documents